1

Paraspeckles: where long noncoding RNA meets phase separation

Archa H Fox*1,4, Shinichi Nakagawa2, Tetsuro Hirose3 and Charles S. Bond4

1. School of Human Sciences, University of Western Australia, Crawley, Western

Australia, 6009 and Harry Perkins Institute of Medical Research, Nedlands, Western

Australia, Australia

2. Faculty of Pharmaceutical Sciences, Hokkaido University, Hokkaido, Japan

3. Institute for Genetic Medicine, Hokkaido University, Hokkaido, Japan

4. School of Molecular Sciences, University of Western Australia, Crawley, Western

Australia, 6009, Australia

*Correspondence: archa.fox@uwa.edu.au (Archa Fox) Keywords Paraspeckles, long noncoding RNA, membraneless organelles, RNA-protein

interactions, gene regulation

Abstract Long noncoding RNA molecules are some of the newest and least understood

players in gene regulation. Hence we need good model systems with well defined

RNA and protein components. One such system is paraspeckles - protein-rich

nuclear organelles built around a specific long noncoding RNA scaffold. New

discoveries show how paraspeckles are formed through multiple RNA-protein and

protein-protein interactions, some of which involve extensive polymerisation, and

others with multivalent interactions driving phase separation. Once formed,

paraspeckles influence gene regulation through sequestration of component proteins

and RNAs, with subsequent depletion in other compartments. Here we focus on the

dual aspects of paraspeckle structure and function, revealing an emerging role for

these dynamic bodies in a multitude of cellular settings.

2

Main text New insights into paraspeckles have helped us understand both lncRNA structure/function and membraneless organelle formation Dynamic cellular organelles that form without a membrane boundary have long

fascinated scientists. What triggers such structures to form? How do a whole variety

of specific molecules get targeted there? How are such structures maintained in

steady state, yet also rapidly broken down when triggered to do so? The

nucleoplasm is rich in such membraneless organelles, often termed ‘nuclear bodies’.

These bodies regulate gene expression, albeit with mechanisms that have been hard

to pin down. Another area that has seen a surge in interest recently is the field of

long noncoding RNA (lncRNA) biology – this has arisen with the knowledge that the

majority of our genome is transcribed, but not translated, yet we have little idea what

all this RNA is doing. Working out the functions of these lncRNAs, how proteins

interact with them and how they can be manipulated has been the focus of much

high profile research. These two important fields - membraneless organelles and

lncRNA biology - coincide spectacularly in the field of nuclear organization,

particularly in one important model system - the paraspeckle - a nuclear body built on

a specific lncRNA.

Discoveries from the last few years have revealed that paraspeckle formation is

dependent on the lncRNA ‘seed’ to recruit specific component proteins by RNA-

protein interactions, thus forcing a local high concentration at a specific nuclear

location. Once in close proximity and bound to their RNA scaffold, these proteins

oligomerise and recruit additional proteins that use low complexity domains to

mediate a liquid liquid phase separation (LLPS) process and formation of the

paraspeckle. Despite this liquid-like state, there is evidence of exquisite substructure

in paraspeckles: super-resolution fluorescence microscopy reveals distinct ‘core’ and

‘shell’ regions containing specific parts of the lncRNA and distinct proteins. In

addition, we have new understanding about how these lncRNA-seeded bodies

regulate gene expression in a dynamic way in different physiological settings,

3

particularly in cancer and cell stress scenarios. In this review we summarise recent

literature on paraspeckles, focusing on these novel developments.

Setting the scene – what are paraspeckles? Paraspeckles are mammal-specific RNA-protein nuclear bodies that regulate gene

expression. First described in 2002 as nuclear foci enriched in the marker protein

PSPC1 (paraspeckle component protein 1) [1], paraspeckles are now known to

contain over 40 different proteins (generally ubiquitously expressed RNA binding

proteins) and one architectural long noncoding RNA, NEAT1 (nuclear paraspeckle

assembly transcript 1, see Table 1). Some component proteins mediate critical RNA-

protein or protein-protein interactions essential for paraspeckle formation (marked in

Table 1). In addition to these structural components there are several different

molecules targeted to paraspeckles, such as various types of RNAs, as well as

proteins that are not required to build paraspeckles, but nevertheless may be

regulated by them.

Paraspeckles of varying size and abundance are found in the nucleus of most

cultured cells, including primary and transformed cell lines, but are absent from

embryonic stem cells (ESC) and induced pluripotent stem cells (iPSC) [2].

Paraspeckles have been detected in many different tissues, however, prominent

paraspeckles are found only in a sub-population of cells in murine tissues, such as

luteal cells in the ovary and cells at the tip of gut epithelium [3].

Although paraspeckles were first defined as sites of enrichment of PSPC1,

subsequent siRNA knockdown experiments showed PSPC1 was in fact dispensable

for paraspeckle formation [4]. However, two other closely-related proteins, NONO

and SFPQ, are absolutely required for paraspeckle formation, along with several

other proteins: HNRNPK, DAZAP1, FUS, RBM14, HNRNPH3 – knockdown of any of

these proteins in HeLa cells results in paraspeckle ablation ([5], table 1). Given this

seeming minor role for the original paraspeckle marker protein, here we propose a

definitive definition of a paraspeckle: a nuclear body in which one of the essential

paraspeckle proteins co-localises with the longer isoform of NEAT1.

The longer isoform of NEAT1 is the structural backbone of paraspeckles

4

The long non coding RNA (lncRNA), NEAT1, is an essential structural component of

paraspeckles, forming a scaffold for protein binding [2, 4, 6, 7]. The requirement of

NEAT1 for paraspeckle formation was one of the first concrete examples of a

functional lncRNA at a time when controversy over their functionality was still raging.

As such, paraspeckles have now become one of the most well studied lncRNA-

protein complexes in the cell.

Two transcripts are generated by RNA Polymerase II from the NEAT1 gene:

NEAT1_1 (3.7kb in humans, 3kb in mice) completely overlaps the 5’ end of NEAT1_2

(23kb in humans, 20kb in mice, Figure 1). Both transcripts contain a single exon and

are nucleus-retained. NEAT1_1 is polyadenylated in a manner akin to a typical

mRNA, whereas NEAT1_2 has an unusual triple helix structure at the 3’ end that has

been used as a characteristic feature for lncRNAs of this class [8]. It is the

transcription of NEAT1_2 that is critically important for paraspeckles, indeed

NEAT1_2 forms the paraspeckle ‘architectural’ backbone [3, 5, 9, 10]. Paraspeckles

fail to form in NEAT1 knockout mice, and only overexpression of NEAT1_2 can

rescue paraspeckle formation in murine embryonic fibroblasts from these mice [5].

Given its central role in paraspeckle structure, it is important to know how NEAT1_2

is made (Figure 1). In vitro and cellular approaches demonstrated that transcription of

NEAT1_2 results from competition between the cleavage factors that would

otherwise cleave and polyadenylate NEAT1_1 being prevented from doing so by the

essential paraspeckle protein HNRNPK [5]. Mutation of the NEAT1_1 polyA site with

genome editing can also mimic this effect [11]. In the absence of transcriptional

termination and cleavage to make NEAT1_1, RNA Polymerase II continues to

transcribe along the DNA, generating the longer NEAT1_2 transcript. Beyond this

initial NEAT1_2 based seeding event, active transcription of NEAT1_2 is required for

paraspeckle formation and maintenance, suggesting additional trans acting factors

that sense NEAT1_2 production [9]. Additional questions remain as to the role of

NEAT1_1 in paraspeckle biology, with new data suggesting it may have a

paraspeckle independent role [11]. Given these distinctions in NEAT1 isoforms, we

urge researchers to consider these differences when designing their experiments and

not assume that NEAT1_1 overexpression, or knockdown in the absence of

NEAT1_2, necessarily influences paraspeckles.

5

Paraspeckles are chains of spheroids, each with a core and shell structure

Paraspeckles are readily labelled with specific marker proteins and RNAs using

fluorescent microscopy, and under the EM are also distinguishable as electron dense

structures in the inter-chromatin space. Despite initially being observed as punctate

foci, EM and super-resolution microscopy has revealed paraspeckles are actually

elongated prolate structures with a uniform width of approximately 360 nm in human

cells and slightly less in murine cells, but with length varying up to 1-2 microns [10].

The number and length of paraspeckles is directly proportional to NEAT1_2

transcription, with more and longer paraspeckles apparent as NEAT1_2 levels

increase [12].

EM and super-resolution microscopy have been used to reveal that paraspeckles are

composed of fused chains of spherical subcompartments each of which have a

distinct core-shell molecular organisation (Figure 2). Fluorescent In situ hybridisation

(FISH) against different regions of NEAT1_2 in the super-resolution microscope

revealed that the 5’ and 3’ ends of this 23,000 nucleotide RNA are clearly confined to

the shell of the paraspeckle, whereas the central region of the RNA is located in the

core [13]. By combining differently coloured probes to either the 3’ or 5’ ends of

NEAT1_2, it was shown that the ends do not individually intermix, but are rather

separated into bundles [13]. Immunofluorescence showed several proteins localised

into distinct zones of the paraspeckle: the essential proteins NONO, SFPQ, FUS are

located in the core, with RBM14 and BRG1 in patches, whereas TDP43 is located in

the shell with the 5’ and 3’ ends of NEAT1 ([13] Figure 2). Although such

ultrastructural insights are largely descriptive, they do provide a critical framework for

studies that aim to understand how this substructure is established in paraspeckle

biogenesis.

How do paraspeckles form?

Paraspeckle biogenesis starts with transcription of NEAT1_2, with paraspeckles

forming close to the NEAT1 gene and often appearing clustered near this locus [6].

Once NEAT1_2 is made, the essential paraspeckle proteins NONO and SFPQ

rapidly bind to form and stabilise a minimal NEAT1_2 RNP particle that is an

intermediate in paraspeckle formation ([13], Figure 1). It is not yet known precisely

which RNA sequence or structure in NEAT1_2 is responsible for NONO/SFPQ

6

binding, but several CLIP (Cross-linking and immunoprecipitation) datasets indicate

extensive binding of the proteins along the length of NEAT1_2 [14]. Again, although

the precise RNA binding modalities are not known, NONO and SFPQ both contain

classical RRM domains that are important in paraspeckle targeting [15]. It is

hypothesised that initial RNA-protein interaction seeding events then trigger

polymerisation of NONO and SFPQ along the RNA. Indeed, coiled-coil domains in

NONO and SFPQ form extensive polymers that are important for the integrity of

paraspeckles [16], with contribution from a conserved charged single charged helix

region that may be involved in the molecular spacing of these proteins along the

NEAT1_2 RNA [17] and low complexity domains that may be involved in protein-

protein interactions. This initial protein binding is thought to be important for

NEAT1_2 stability as both isoforms of NEAT1 have a relatively short half life [18, 19].

Additional core proteins - principally FUS - are required to knit together these

NONO/SFPQ/NEAT1_2 RNPs into a mature paraspeckle [13, 20], Figure 1).

Interestingly, RBM14 and the swi/snf component BRG1 are both found in patches

bridging the core and shell regions of the paraspeckle and these two proteins are

both required for essential additional protein-protein interactions to form

paraspeckles [21, 22]. However, BRG1 does not require its ATPase catalytic site for

this structural role. In terms of stoichiometry, recent quantitative analysis revealed the

presence of ~50 NEAT1_2 molecules in a single paraspeckle particle [23], but the

stoichiometry of different paraspeckle proteins is yet to be determined.

The proteins that build paraspeckles have RNA binding and polymerisation

propensity that is also observed in other contexts. For example, NONO and SFPQ

can be rapidly recruited to artificial scaffolds such as transfected antisense

oligonucleotides to form ‘pseudo-paraspeckles’, devoid of NEAT1 [24]. However, it

remains to be tested if the ordered core-shell structures of paraspeckle proteins is

also found in the protein assembly induced by these artificial scaffolds.

Liquid liquid phase separation is critical for paraspeckle formation

Paraspeckles actually share many features with cytoplasmic stress granules, another

type of membraneless organelle. Paraspeckles and stress granules both contain

common component proteins [25], become more abundant with stress, seem to

7

function through molecular sequestration (more on this below) and both have distinct

subsets of molecules found either within the core, or shell regions [13, 26]. Our

understanding of the type of molecular interactions driving formation of such

membraneless organelles has been changing with the appreciation that these

structures may be more liquid than solid, indeed they could be undergoing liquid

liquid phase separation (LLPS). Exciting work on protein low complexity domains -

domains previously thought to be flexible linkers or regions incapable of folding -

show that these regions can mediate LLPS [27]. The rapid aggregation of these

proteins can result in a continuum of states, from monomers, to liquids, to fibrillar

amyloid-like structures that contain cross beta interactions [28]. However, unlike

pathological amyloid, such fibrillar structures are capable of being broken down.

These fibrillar structures can form hydrogels in vitro [29]. Interestingly, nuclear bodies

have long been observed to have liquid like properties involving fusion and splitting

[30].

The sub-class of low complexity domain that is responsible for fibril formation is the

‘prion-like domain’, named due to the similarity to the yeast prion protein. Found in

almost half of the paraspeckle proteins (Table 1), prion-like domains are enriched in

uncharged polar amino acids such as asparagine, glutamine, tyrosine, as well as

glycine [31]. Of the proteins responsible for protein-protein interactions in the

paraspeckle interactome, two thirds contain prion-like domains, indicating that a

network of prion-like domain mediated interactions may be extensive within

paraspeckles [21]. A combination of in vitro and cell experiments were used to show

that two essential paraspeckle proteins, RBM14 and FUS, both require an intact

prion-like domain to be targeted to, and effectively build, paraspeckles [13, 21, 23]. In

these experiments, mutants of FUS and RBM14 with Y to S substitutions in their

prion like domains (mutations that compromise the ability of these domains to form

fibrils and hydrogels) were incapable of forming paraspeckles [21]. Additional cellular

studies on FUS, TDP43 and HnRNPA1 all attest to the importance of their prion-like

domains for liquid phase separation [32-34].

Remarkably, the association of NEAT1_2 with the paraspeckle proteins is so

extensive that the RNA is not easily extracted with conventional RNA solubilisation

methods such as Trizol. Heating and needle-shearing is required to extract the RNA

8

fully from the protein phase, leading to an up to 10 fold increase in NEAT1_2

extraction. Indeed, this ‘semi-extractability’ of NEAT1_2 is dependent on the prion

like domain of FUS, as cells expressing FUS truncation mutants that lack the prion-

like domain have freely extractable NEAT1_2 [23]. It is therefore intriguing to

speculate that the structural basis of this semi-extractability of NEAT1_2 may involve

some form of FUS-driven phase separation.

What is the molecular mechanism of paraspeckle function?

Once paraspeckles form, what is their role in the cell? As yet, no distinct catalytic

activity has been demonstrated to occur within the paraspeckle. The main molecular

mechanisms identified for paraspeckle function relate to sequestration of molecules,

be it RNA or proteins (summarised in Figure 3).

Paraspeckles trap certain RNA species in the nucleus Beyond NEAT1, other types of RNA species can localise to paraspeckles and

potentially be regulated by them. The best characterised class of such nucleus-

retained RNAs contain double-stranded RNA structures formed through transcribed

inverted repeat motifs [35]. This nuclear retention mechanism is driven by NONO and

SFPQ binding to the dsRNA structure in the RNA. Recent studies have shown that

methylation of the coiled-coil region of NONO is important for regulating its ability to

bind these dsRNAs [36]. How might this nuclear retention mechanism be used in cell

physiology? One example has emerged for paraspeckles having a role in the post-

transcriptional control of gene regulation in a circadian context. Rhythmic expression

of paraspeckle/NEAT1_2 levels is observed in pituitary cells and knockdown of

NEAT1_2 prevents the rhythmic expression of inverted repeat-fused reporters [37].

Paraspeckles quantitatively sequester their protein components

As NEAT1_2 levels increase, paraspeckles elongate ([12], Figure 3). A consequence

of increasing paraspeckle length is that a greater amount of constituent paraspeckle

protein is required to build the structures, with a concomitant reduction in

nucleoplasmic protein availability. This concept has been well defined for the

paraspeckle protein SFPQ. As paraspeckle elongate, requiring more SFPQ, the

relative levels of SFPQ decrease in the nucleoplasm. The consequence of SFPQ

nucleoplasmic depletion is alteration in its direct target genes. For example, SFPQ

9

normally represses IL-8, an immune responsive gene, or activates the ADARB2 gene

[12, 38]. For paraspeckles, SFPQ is the only molecule to have this detailed study

performed, but it is presumed that all paraspeckle proteins will be regulated by

changing paraspeckle size in a similar manner. Other nuclear bodies are known to

act with a similar subnuclear sequestration mechanism. For example, nuclear bodies

seeded by the SPA lncRNA (with a 5' snoRNA cap and 3' polyadenylation) that trap

and deplete other specific RNA binding proteins from the nucleoplasm [39].

New paraspeckle roles in transcription and miRNA processing

A further possibility exists for paraspeckles to directly dock onto the transcription start

sites of actively transcribed genes. A CHART (capture hybridization analysis of RNA

targets) experiment showed that NEAT1 RNA crosslinked to the transcription start

sites of active genes in a breast cancer cell line, but the probes targeted both

NEAT1_1 and NEAT1_2 therefore it is unclear if this is a paraspeckle specific

activity, or NEAT1_1 driven activity, being revealed [40]. Further, SFPQ and NONO

were recently found to have activity in pri-miRNA processing, enhancing the

processing of introns containing pri-miRNA, leading to increasing mature miRNAs

[14]. By sequestration of SFPQ and NONO into paraspeckles, along with some of the

miRNA processing factors, the miRNA processing is positively influenced [14]. It is

thought in this instance that the association of the proteins into the paraspeckle is

positively enhancing this pri-miRNA processing capability of the proteins by providing

an efficient platform for their interactions (Figure 3).

Paraspeckles play a role in a variety of developmental and disease scenarios

In which biological settings might paraspeckles play a role in regulating gene

expression? The paraspeckle proteins can be found both inside and outside

paraspeckles, therefore study of paraspeckle protein function cannot be seen as akin

to paraspeckle function. In contrast, NEAT1_2 is strictly only found within

paraspeckles, and NEAT1_2 levels can be considered a proxy for paraspeckle

appearance. Thus observation and manipulation of NEAT1_2 is a route to functional

studies of paraspeckles. Here we have distilled functional information about

paraspeckles, focusing on studies that have distinguished the NEAT1 isoforms

and/or conducted microscopic imaging to detect paraspeckles in different biological

settings.

10

Paraspeckles are important in female reproduction

Paraspeckles and Neat1 are not essential for mammalian development under normal

conditions - it was originally reported that mice lacking Neat1 were viable [3].

Subsequent analyses revealed that a subset of female Neat1-/- mice had impaired

formation of the transient ovarian secretory gland, the corpus luteum, that supports

pregnancy (in wildtype mice the corpus luteum has large, prominent paraspeckles).

Lack of a formation of pregnant corpus luteum leads to infertility, smaller litters and

fewer viable pregnancies for knockout females [41]. There may be an evolutionary

conserved function for paraspeckles in secretory structures of the female

reproductive system: the marsupial opossum has distinct paraspeckles in the uterine

gland, however their functional significance is not known [42]. Neat1 knockout mice

also exhibit deficiencies in mammary epithelial cell proliferation leading to a lactation

defect in the mothers and low pup survival [43]. Curiously, some Neat1 knockout

phenotypes have a stochastic effect, for example the corpus luteum will form in

some, but not all female knockouts. These stochasitc effects, along with in vitro

studies indicating stress induction of paraspeckles through environmental triggers

has led to a theme that NEAT1/paraspeckles exert their functional roles when cells

are stressed.

Paraspeckles increase with viral infection

NEAT1_2 levels have been observed increasing in different cell types following

infection with many different RNA viruses, including Japanese encephalitis, rabies

[44], HIV [45, 46], influenza [38] and hantaan [47], as well as the DNA-encoded

herpes simplex virus [48]. In most cases the evidence suggests that

NEAT1_2/paraspeckles increase as a cell defence mechanism. More abundant

paraspeckles lead to increased sequestration of the paraspeckle proteins, thereby

preventing the hijacking of these proteins by the virus [48, 49], or triggering gene

regulatory changes that help the cell [38, 47].

Are paraspeckles friend or foe in cancer?

Whilst many studies implicate the molecule NEAT1 in cancer progression, there are

few studies directly connecting paraspeckles, or NEAT1_2, to cancer biology. A meta

analysis of NEAT1 in cancer revealed a complex and varying pattern of expression

11

revealing that NEAT1 can be either upregulated, downregulated, or unchanged in

tumor compared to normal tissue, depending on different studies and cancer models,

with most studies not differentiating between NEAT1 isoforms [50]. Despite this, the

possibility for a role for paraspeckles in cancer is gaining more credence. The NEAT1

gene, including the region encoding NEAT1_2, contains a significantly large number

of mutations in tumors from liver cancer patients [51], and the NEAT1 promoter was

recently found to be a hotspot for mutations in breast cancer [52]. Overall several

different cancer subtypes show increasing numbers of mutations in NEAT1_2 [53]. A

common theme has emerged of NEAT1/paraspeckles being induced by stress in pre-

neoplastic, or tumor cells. In many such cases, however, a function is only apparent

when stress is applied to the cell.

In many settings paraspeckles seem to have oncogenic roles. Hypoxia triggers

increased NEAT1_2 transcription through hypoxia-inducible factor (HIF) with a

resultant increase in paraspeckles in breast cancer cell lines [54]. In prostate cancer,

increased levels of NEAT1_1 as well as NEAT1_2/paraspeckles are associated with

advanced forms of the disease [55]. In this case, NEAT1 transcription is stimulated

by the estrogen receptor (ER) and together NEAT1 and ER promote transcription of

a circuit of proliferative genes. Overexpression of NEAT1_1 increased active

epigenetic marks at target genes and triggered proliferation, with loss of both NEAT1

isoforms reducing proliferation [55]. In skin fibroblasts, p53 induces

NEAT1_2/paraspeckles when pre-neoplastic cells transform into tumors, either in

genetic or topical skin cancer mouse models, and loss of NEAT1 prevents this

transformation, suggesting NEAT1 is a potent oncogene [56]. In that study NEAT1

and paraspeckles prevented replication stress, aiding the DNA damage response,

although the molecular mechanism was not elucidated.

In opposition to observations of paraspeckles as oncogenic, other studies find

paraspeckles have tumor suppressor roles. P53 induction of NEAT1_1 and

NEAT1_2/paraspeckles in cells derived from RAS-driven pancreatic cancer mouse

models prevented transformation in vitro and also prevented pancreatic cancer

progression in murine models [57]. Interestingly, overexpression of NEAT1_1 in p53-

/- cells partially restored the p53-mediated suppression of transformation, suggesting

that in this context, NEAT1_1 may be playing a role in tumor suppression, rather than

12

paraspeckles per se. In different types of blood cancers, NEAT1 levels correlate

inversely with proliferation [58]. In colorectal cancer cell lines, inhibition of NEAT1_2

enhanced proliferation, suggesting paraspeckles could be acting as a tumor

suppressor in this context [59].

Paraspeckles in the nervous system

Some areas of the mouse brain show increased NEAT1_2 levels following seizures

[60] and experiments in iPSC-derived neurons indicate that NEAT1 function, but not

necessarily paraspeckles per se, may regulate hyperexcitability and have links to

epilepsy [61]. Paraspeckles may also have a role in motor neuron disease (MND,

also known as amyotropic lateral sclerosis, or ALS). Several paraspeckle proteins

are encoded by genes mutated in familial forms of ALS (Table 1). When ALS patient

derived mutants of FUS are expressed in cell models, cytoplasmic aggregates form

that can also be enriched for NONO and SFPQ [20]. Motor neurons do not normally

express Neat1_2, but its expression is upregulated in early-phase degenerating

motor neurons of ALS patients [62]. Overall, however, the physiological significance

of the formation of paraspeckles during ALS pathogenesis remains to be elucidated.

Concluding Remarks and Future Perspectives In the 15 years since their discovery, paraspeckles have flourished from obscure

nuclear body to important lncRNA-protein model system with novel roles in gene

regulation in different disease contexts. By studying how paraspeckles form we are

learning more about how nuclear proteins use their RNA binding and protein-protein

aggregation propensity to form such membraneless organelles. With many questions

still remaining about paraspeckle biology (see ‘outstanding questions’) we are

confident that the field will continue to grow, providing a rich source of novel

understanding into gene regulation.

Figure legends Figure 1. Biogenesis of a paraspeckle. Generation of the overlapping NEAT1

isoforms (A) is first triggered by activation of the NEAT1 promoter and recruitment of

RNA polymerase II. Once the sequence corresponding to NEAT1_1 is transcribed, a

canonical polyA site at the NEAT1_1 3’end recruits cleavage and polyadenylation

13

factors that will generally terminate transcription (B). In some instances, this cleavage

and polyadenylation is prevented from occurring by the inhibition of these processes

by HNRNPK, which binds to the same region of NEAT1_1 (C). When this happens,

RNA Polymerase II continues to transcribe through into the NEAT1_2 region,

generating the longer 23kb NEAT1_2 transcript that contains the specific binding

sites for NONO and SFPQ (D) that bind and polymerise along the RNA forming

NEAT1_2/NONO/SFPQ RNA-protein bundles (E). These bundles are then fused into

a microscopically visible mature paraspeckle by FUS, with an activity dependent on

the FUS prion like low complexity domain (F). (G) A confocal image of a HeLa cell

stained with an antibody to the paraspeckle protein, NONO (green), with the

fluorescent signal overlaid on a bright-field image of the cell. Arrows indicate

paraspeckles. Scale bar 5 µm.

Figure 2. Paraspeckle substructure contains core and shell zones.

Fluorescence micrograph on the left shows a super-resolution/structured illumination

image of paraspeckles from murine corpus luteal cells with the shell (green)

highlighted by FISH with combined probes recognising the 5’ and 3’ ends of Neat1_2,

and the core region (magenta) highlighted by FISH with probes to the middle

sequences of Neat1_2. Both single mature paraspeckles, as well as chains of

paraspeckle units, are depicted. Scale bar 500 nm. The model on the right

summarises super-resolution imaging, depicting the relative locations of different

paraspeckle component proteins and RNA regions to either the core or shell zones of

the paraspeckle. Image and model adapted from [13].

Figure 3. Paraspeckles regulate gene expression by sequestration and enhancing the efficiency of paraspeckle protein function. The steady state is

depicted in the centre, with paraspeckle proteins and targeted structured RNAs found

both inside paraspeckles, as well as elsewhere in the nucleus, including regulating

gene transcription at promoters. When the paraspeckle proteins are organised inside

paraspeckles there is evidence for enhanced processing of pri-miRNA transcripts.

The right panel depicts that happens when cells become stressed, or NEAT1_2 is

increased. In this instance, paraspeckles elongate, increasing the sequestration of

paraspeckle proteins, with the pool of nucleoplasmic paraspeckle protein diminishing,

14

resulting in a change in transcription of target genes. The left panel shows what

happens when NEAT1_2 levels decrease, either in a particular stage of the circadian

cycle, in cells that do not express NEAT1_2, or when NEAT1_2 is artificially reduced.

In this instance, paraspeckle proteins and regulated RNAs are free to bind to target

gene promoters, or, in the case of regulated RNAs, to be exported from the nucleus

and used as templates for translation. The efficiency of pri-miRNA processing by

paraspeckle proteins is diminished.

Acknowledgements We are grateful to the members of our laboratories for helpful reading of this

manuscript and apologize to those whose work we could not cite due to space

limitations. This work was supported by the Australian Research Council grant

DP160102435 to AHF and CSB.

15

Table 1 – Paraspeckle proteins and RNAs gene name importance in

paraspeckle formation

prion like domain1

ALS2- mutat

ion

LLPS3 link

paraspeckle zone4

[13]

reference

Paraspeckle proteins AHDC1 dispensible no [5] AKAP8L dispensible yes [5] CELF6 n. d. no [5] CIRBP dispensible no [5] CPSF5 dispensible no [5] CPSF6 dispensible no [5] CPSF7 important no [5]

DAZAP1 essential yes [5] DLX3 n. d. yes [5]

EWSR1 dispensible yes yes [5] [63] FAM98A important yes [5]

FIGN important yes [5] FUS essential yes yes yes core [64][5]

HNRNPA1 important yes yes yes [65][5] HNRNPA1L2 n. d. yes [5]

HNRNPF n. d. no [5] HNRNPH1 n. d. yes [5] HNRNPH3 essential yes [5] HNRNPK essential no [5] HNRNPR important yes [5]

HNRNPUL1 important yes [5] MEX3A n. d. no [5] NONO essential yes core [15]

PCED1A important no [5] PSPC1 dispensible yes core [5] RBM3 dispensible yes [5]

RBM4B dispensible no [5] RBM7 dispensible no [5]

RBM12 important yes [5] RBM14 essential yes yes patch [15] RBMX dispensible no [5] RUNX3 dispensible yes [5] SFPQ essential yes yes core [66] [15]

SMARCA4 (BRG1)

essential no patch [22]

SRSF10 important no [5] SS18L1 n. d. yes yes [67][5] TAF15 important yes yes [63] [5] TDP43 n.d. yes yes shell [62]

UBAP2L dispensible yes [5] ZC3H6 dispensible yes [5]

Paraspeckle RNAS

NEAT1 essential N/A 5’+3’ shell, middle core

[2, 4, 6, 7]

IR-containing dispensible N/A [35]

16

RNAs AG rich RNAs dispensible N/A shell [13]

n.d. – not determined N/A – not applicable 1 A type of low complexity domain rich in polar and small amino acids (Gly, Ala, Ser, Pro, Asn, Gln, Tyr) implicated in forming fibrillar higer-order aggregates. 2 Amyotrophic lateral sclerosis, also known as Motor Neuron Disease 3 partition of components of molecular mixtures into distinct demixed phases, e.g. oil and water. In the cell, many membraneless organelle display liquid behaviour suggesting that they are demix liquids. 4 The paraspeckle zones relate to the super-resolution imaging of paraspeckles as performed in [13]

17

References

1.Fox,A.H.etal.(2002)Paraspeckles:anovelnucleardomain.CurrBiol12(1),13-25.2.Chen,L.L.andCarmichael,G.G.(2009)AlterednuclearretentionofmRNAscontaininginvertedrepeatsinhumanembryonicstemcells:functionalroleofanuclearnoncodingRNA.MolCell35(4),467-78.3.Nakagawa,S.etal.(2011)Paraspecklesaresubpopulation-specificnuclearbodiesthatarenotessentialinmice.JCellBiol193(1),31-9.4.Sasaki,Y.T.etal.(2009)MENepsilon/betanoncodingRNAsareessentialforstructuralintegrityofnuclearparaspeckles.ProcNatlAcadSciUSA106(8),2525-30.5.Naganuma,T.etal.(2012)Alternative3'-endprocessingoflongnoncodingRNAinitiatesconstructionofnuclearparaspeckles.EMBOJ31(20),4020-34.6.Clemson,C.M.etal.(2009)AnarchitecturalroleforanuclearnoncodingRNA:NEAT1RNAisessentialforthestructureofparaspeckles.MolCell33(6),717-26.7.Sunwoo,H.etal.(2009)MENepsilon/betanuclear-retainednon-codingRNAsareup-regulateduponmuscledifferentiationandareessentialcomponentsofparaspeckles.GenomeRes19(3),347-59.8.Zhang,B.etal.(2017)IdentificationandCharacterizationofaClassofMALAT1-likeGenomicLoci.CellRep19(8),1723-1738.9.Mao,Y.S.etal.(2011)Biogenesisandfunctionofnuclearbodies.TrendsGenet27(8),295-306.10.Souquere,S.etal.(2010)HighlyorderedspatialorganizationofthestructurallongnoncodingNEAT1RNAswithinparaspecklenuclearbodies.MolBiolCell21(22),4020-7.11.Li,R.etal.(2017)FunctionaldissectionofNEAT1usinggenomeeditingrevealssubstantiallocalizationoftheNEAT1_1isoformoutsideparaspeckles.RNA23(6),872-881.12.Hirose,T.etal.(2014)NEAT1longnoncodingRNAregulatestranscriptionviaproteinsequestrationwithinsubnuclearbodies.MolBiolCell25(1),169-83.13.West,J.A.etal.(2016)Structural,super-resolutionmicroscopyanalysisofparaspecklenuclearbodyorganization.JCellBiol214(7),817-30.14.Jiang,L.etal.(2017)NEAT1scaffoldsRNA-bindingproteinsandtheMicroprocessortogloballyenhancepri-miRNAprocessing.NatStructMolBiol24(10),816-824.15.Fox,A.H.etal.(2005)P54nrbformsaheterodimerwithPSP1thatlocalizestoparaspecklesinanRNA-dependentmanner.MolBiolCell16(11),5304-15.16.Lee,M.etal.(2015)ThestructureofhumanSFPQrevealsacoiled-coilmediatedpolymeressentialforfunctionalaggregationingeneregulation.NucleicAcidsRes43(7),3826-40.17.Dobson,L.etal.(2015)Aconservedchargedsinglealpha-helixwithaputativestericroleinparaspeckleformation.RNA21(12),2023-9.18.Clark,M.B.etal.(2012)Genome-wideanalysisoflongnoncodingRNAstability.GenomeRes22(5),885-98.19.Tani,H.etal.(2012)Genome-widedeterminationofRNAstabilityrevealshundredsofshort-livednoncodingtranscriptsinmammals.GenomeRes22(5),947-56.

18

20.Shelkovnikova,T.A.etal.(2014)CompromisedparaspeckleformationasapathogenicfactorinFUSopathies.HumMolGenet23(9),2298-312.21.Hennig,S.etal.(2015)Prion-likedomainsinRNAbindingproteinsareessentialforbuildingsubnuclearparaspeckles.JCellBiol210(4),529-39.22.Kawaguchi,T.etal.(2015)SWI/SNFchromatin-remodelingcomplexesfunctioninnoncodingRNA-dependentassemblyofnuclearbodies.ProcNatlAcadSciUSA112(14),4304-9.23.Chujo,T.etal.(2017)Unusualsemi-extractabilityasahallmarkofnuclearbody-associatedarchitecturalnoncodingRNAs.EMBOJ36(10),1447-1462.24.Shen,W.etal.(2014)PhosphorothioateoligonucleotidescandisplaceNEAT1RNAandformnuclearparaspeckle-likestructures.NucleicAcidsRes42(13),8648-62.25.Jain,S.etal.(2016)ATPase-ModulatedStressGranulesContainaDiverseProteomeandSubstructure.Cell164(3),487-98.26.Wheeler,J.R.etal.(2017)Isolationofyeastandmammalianstressgranulecores.Methods126,12-17.27.Kaganovich,D.(2017)ThereIsanInclusionforThat:MaterialPropertiesofProteinGranulesProvideaPlatformforBuildingDiverseCellularFunctions.TrendsBiochemSci42(10),765-776.28.Kato,M.andMcKnight,S.L.(2017)Cross-betaPolymerizationofLowComplexitySequenceDomains.ColdSpringHarbPerspectBiol9(3).29.Kato,M.etal.(2012)Cell-freeformationofRNAgranules:lowcomplexitysequencedomainsformdynamicfiberswithinhydrogels.Cell149(4),753-67.30.Platani,M.etal.(2000)InvivoanalysisofCajalbodymovement,separation,andjoininginlivehumancells.JCellBiol151(7),1561-74.31.Lancaster,A.K.etal.(2014)PLAAC:awebandcommand-lineapplicationtoidentifyproteinswithprion-likeaminoacidcomposition.Bioinformatics30(17),2501-2.32.Patel,A.etal.(2015)ALiquid-to-SolidPhaseTransitionoftheALSProteinFUSAcceleratedbyDiseaseMutation.Cell162(5),1066-77.33.Schmidt,H.B.andRohatgi,R.(2016)InVivoFormationofVacuolatedMulti-phaseCompartmentsLackingMembranes.CellRep16(5),1228-1236.34.Shin,Y.etal.(2017)SpatiotemporalControlofIntracellularPhaseTransitionsUsingLight-ActivatedoptoDroplets.Cell168(1-2),159-171e14.35.Prasanth,K.V.etal.(2005)RegulatinggeneexpressionthroughRNAnuclearretention.Cell123(2),249-63.36.Hu,S.B.etal.(2015)ProteinargininemethyltransferaseCARM1attenuatestheparaspeckle-mediatednuclearretentionofmRNAscontainingIRAlus.GenesDev29(6),630-45.37.Torres,M.etal.(2016)CircadianRNAexpressionelicitedby3'-UTRIRAlu-paraspeckleassociatedelements.Elife5.38.Imamura,K.etal.(2014)LongnoncodingRNANEAT1-dependentSFPQrelocationfrompromoterregiontoparaspecklemediatesIL8expressionuponimmunestimuli.MolCell53(3),393-406.39.Wu,H.etal.(2016)UnusualProcessingGeneratesSPALncRNAsthatSequesterMultipleRNABindingProteins.MolCell64(3),534-548.40.West,J.A.etal.(2014)ThelongnoncodingRNAsNEAT1andMALAT1bindactivechromatinsites.MolCell55(5),791-802.41.Nakagawa,S.etal.(2014)ThelncRNANeat1isrequiredforcorpusluteumformationandtheestablishmentofpregnancyinasubpopulationofmice.Development141(23),4618-27.

19

42.Cornelis,G.etal.(2016)FunctionalconservationofthelncRNANEAT1intheancestrallydivergedmarsupiallineage:EvidenceforNEAT1expressionandassociatedparaspeckleassemblyduringlategestationintheopossumMonodelphisdomestica.RNABiol13(9),826-36.43.Standaert,L.etal.(2014)ThelongnoncodingRNANeat1isrequiredformammaryglanddevelopmentandlactation.RNA20(12),1844-9.44.Saha,S.etal.(2006)Identificationandcharacterizationofavirus-induciblenon-codingRNAinmousebrain.JGenVirol87(Pt7),1991-5.45.Zhang,Q.etal.(2013)NEAT1longnoncodingRNAandparaspecklebodiesmodulateHIV-1posttranscriptionalexpression.MBio4(1),e00596-12.46.Jin,C.etal.(2016)DetectionofthelongnoncodingRNAsnuclear-enrichedautosomaltranscript1(NEAT1)andmetastasisassociatedlungadenocarcinomatranscript1intheperipheralbloodofHIV-1-infectedpatients.HIVMed17(1),68-72.47.Ma,H.etal.(2017)TheLongNoncodingRNANEAT1ExertsAntihantaviralEffectsbyActingasPositiveFeedbackforRIG-ISignaling.JVirol91(9).48.Wang,Z.etal.(2017)NEAT1modulatesherpessimplexvirus-1replicationbyregulatingviralgenetranscription.CellMolLifeSci74(6),1117-1131.49.Budhiraja,S.etal.(2015)MiningthehumancomplexomedatabaseidentifiesRBM14asanXPO1-associatedproteininvolvedinHIV-1Revfunction.JVirol89(7),3557-67.50.Xiong,D.D.etal.(2017)TheclinicalvalueoflncRNANEAT1indigestivesystemmalignancies:Acomprehensiveinvestigationbasedon57microarrayandRNA-seqdatasets.Oncotarget8(11),17665-17683.51.Fujimoto,A.etal.(2016)Whole-genomemutationallandscapeandcharacterizationofnoncodingandstructuralmutationsinlivercancer.NatGenet48(5),500-9.52.Rheinbay,E.etal.(2017)Recurrentandfunctionalregulatorymutationsinbreastcancer.Nature547(7661),55-60.53.Lanzos,A.etal.(2017)DiscoveryofCancerDriverLongNoncodingRNAsacross1112TumourGenomes:NewCandidatesandDistinguishingFeatures.SciRep7,41544.54.Choudhry,H.etal.(2015)TumorhypoxiainducesnuclearparaspeckleformationthroughHIF-2alphadependenttranscriptionalactivationofNEAT1leadingtocancercellsurvival.Oncogene34(34),4482-90.55.Chakravarty,D.etal.(2014)Theoestrogenreceptoralpha-regulatedlncRNANEAT1isacriticalmodulatorofprostatecancer.NatCommun5,5383.56.Adriaens,C.etal.(2016)p53inducesformationofNEAT1lncRNA-containingparaspecklesthatmodulatereplicationstressresponseandchemosensitivity.NatMed22(8),861-8.57.Mello,S.S.etal.(2017)Neat1isap53-induciblelincRNAessentialfortransformationsuppression.GenesDev31(11),1095-1108.58.Zeng,C.etal.(2014)Inhibitionoflongnon-codingRNANEAT1impairsmyeloiddifferentiationinacutepromyelocyticleukemiacells.BMCCancer14,693.59.Wu,Y.etal.(2015)Nuclear-enrichedabundanttranscript1asadiagnosticandprognosticbiomarkerincolorectalcancer.MolCancer14,191.60.Bluthgen,N.etal.(2017)ProfilingtheMAPK/ERKdependentandindependentactivityregulatedtranscriptionalprogramsinthemurinehippocampusinvivo.SciRep7,45101.61.Barry,G.etal.(2017)Thelongnon-codingRNANEAT1isresponsivetoneuronalactivityandisassociatedwithhyperexcitabilitystates.SciRep7,40127.

20

62.Nishimoto,Y.etal.(2013)Thelongnon-codingRNAnuclear-enrichedabundanttranscript1_2inducesparaspeckleformationinthemotorneuronduringtheearlyphaseofamyotrophiclateralsclerosis.MolBrain6,31.63.Couthouis,J.etal.(2011)AyeastfunctionalscreenpredictsnewcandidateALSdiseasegenes.ProcNatlAcadSciUSA108(52),20881-90.64.Vance,C.etal.(2009)MutationsinFUS,anRNAprocessingprotein,causefamilialamyotrophiclateralsclerosistype6.Science323(5918),1208-1211.65.Kim,H.J.etal.(2013)Mutationsinprion-likedomainsinhnRNPA2B1andhnRNPA1causemultisystemproteinopathyandALS.Nature495(7442),467-73.66.Thomas-Jinu,S.etal.(2017)Non-nuclearPoolofSplicingFactorSFPQRegulatesAxonalTranscriptsRequiredforNormalMotorDevelopment.Neuron94(2),322-336e5.67.Chesi,A.etal.(2013)ExomesequencingtoidentifydenovomutationsinsporadicALStrios.NatNeurosci16(7),851-5.

Glossary Paraspeckle: micron-sized ribonucleoprotein granule found in the mammalian cell nucleus Long noncoding RNA: An RNA transcript over 200 nucleotides in lengths which has no coding potential. New functional lncRNAs with roles in gene regulation and/or structure (architectural lncRNA) are constantly being discovered. Membraneless organelle: a functional subcellular body which is not bound by a lipid membrane. Without a membrane boundary, the organelle must be held together by interactions between its component molecules. Liquid-liquid phase separation: partition of components of molecular mixtures into distinct demixed phases, e.g. oil and water. In the cell, many membraneless organelle display liquid behaviour suggesting that they are demix liquids. Core-shell structure: A three-dimensional object with regions of varying composition. The core forms the centre of the object which is surrounded by the shell (e.g. watermelon). Protein low complexity domains: Regions of a protein containing a small representation of amino acid types, sometimes as polymers (e.g. polyproline, polyalanine), as tandem-repeats (e.g. Tyr-Xaa-Xaa-Gln repeats) or with identifiable functional motifs (e.g. Arg-Gly-Gly motifs for nucleic acid binding). Prion-like domains: A type of low complexity domain rich in polar and small amino acids (Gly, Ala, Ser, Pro, Asn, Gln, Tyr) implicated in forming fibrillar higer-order aggregates.

Outstanding questions

• What are the elements in the NEAT1_2 RNA that the paraspeckle

proteins recognise and bind? Given relatively low sequence homology

for NEAT1_2 in mammals, is it possible these protein-binding elements

are set by secondary or tertiary structure of NEAT1_2? How do

mutations in NEAT1_2 in cancer cells impact protein binding and

paraspeckle formation?

• NONO and SFPQ function in miRNA processing is enhanced by their

organisation into paraspeckles – will this principle hold true for other

paraspeckle protein activities that could act in trans? Many paraspeckle

proteins have roles in alternative splicing, could such activities also be

enhanced by paraspeckle organisation? It is interesting to speculate as

paraspeckles are found adjacent to nuclear speckles, bodies rich in

splicing factors.

• What mechanism regulates individual paraspeckle length? Which

factors allow the single paraspeckle units to form a longer chain?

• What is the interrelationship between NEAT1_1 and NEAT1_2? Is it

possible that one role for NEAT1_2/paraspeckle production is to

regulate through negative feedback, the otherwise independent

NEAT1_1 molecule that may have powerful oncogenic activity?

• Paraspeckles generally form close to the NEAT1 gene and then can

move elsewhere in the nucleus. But why do even distally located,

mature paraspeckles depend on active transcription of NEAT1 for their

integrity? How does the signal that transcription has ceased get

transmitted to such paraspeckles?

• Are there additional proteins targeted to paraspeckles that are yet to be

discovered?

• What are the key signatures in prion like domains that target them to

paraspeckles, are there other nuclear proteins with such domains that

do not get targeted there? How distinct is one prion like domain from

another?

Trends box

• Paraspeckles have become an important model system for studying

long noncoding RNA-protein interactions in the context of gene

regulation

• Nascent 23,000 nucleotide NEAT1 long noncoding RNA transcripts act

as a seed to recruit nuclear RNA binding proteins and build a

paraspeckle

• Protein domains that mediate Liquid liquid phase separation are

essential for many aspects of paraspeckle formation, including glueing

together individual ribonucleoprotein bundles into a mature

paraspeckle

• Paraspeckle formation is dynamic and triggered by many different cell

stress scenarios including infection and transformation

Chr11q3.1

NEAT1_1

NEAT1_2

AAAAAAtriplehelix

CFIm

HNRNPK

AAAA

NEAT1_2/NONO/SFPQRNP

Single matureparaspeckle

Pol IIPol II Pol II

5'

5'

tRNA-likestructure

5'3'

5'

NONO/SFPQdimers

A B

C

DE

F

+FUS and other factors

FUS

FUS

FUSFUS

FUS

G

Steady State (paraspeckle )NEAT1_2NEAT1_2

translation

eg. circadian cycling eg. stress

pri-miRNA processing

(paraspeckle )

altered transcription

eg. normal

paraspeckle protein paraspeckle-targeted RNA

pri-miRNA paraspeckle

nucleus

cytoplasm